When one moves their hand from one point to another, the brain guides the arm by relying on neural structures that estimate physical dynamics of the task and transform the desired motion into motor commands. If our hand is holding an object, the subtle changes in the dynamics of the arm are taken into account by these neural structures and this is reflected in the altered motor commands. These observations have suggested that in generating motor commands, the brain strongly relies on internal models that predict physical dynamics of the external world. The internal models are learned with practice, and appear to be a fundamental part of voluntary motor control. However, we know very little about which neural structures in the brain are involved in formation of internal models for motor control and how they learn to represent these models. Our aim here is to combine behavioral and mathematical tools to infer how humans learn internal models, and how the process is affected when there is damage to specific motor structures in the brain. We approach the problem by considering a task where physical dynamics of reaching movements are altered. As people practice the task, we ask how did an error that was experienced in a given movement affect subsequent movements? We arrive at a generalization function that mathematically describes how the brain changes the internal model in response to an error. The shape of the generalization function predicts the receptive field of the elements that took part in representing the internal model with respect to movement kinematics. We study these generalization functions in position, velocity, and acceleration space of the arm. Preliminary results demonstrate a remarkable similarity between these behaviorally inferred bases and typical tuning properties of cells in the primary motor cortex. This suggests that tuning properties of these cells might be reflected in human behavior in the way that we learn and generalize patterns of force. We extend the studies to include generalization from one arm to the other. We further extend the studies to movements where dynamics are not dependent only on arm kinematics, but also on other cues: external cues where dynamics are linked to an arbitrary spatial or color cue, internal cues where dynamics depends on position of a movement with in a sequence. We compare how damage to the brain in Huntington's disease vs. cerebellar disease affects this learning. However, motor memories are not static. Their functional properties change within hours after a task is learned. We ask how this change affects the generalization function. Is the brain different in the way it responds to an error after a memory has consolidated vs. early in the learning phase? [unreadable] [unreadable]